virological synapse-mediated spread of human immunodeficiency

JOURNAL OF VIROLOGY, Apr. 2010, p. 3516–3527 Vol. 84, No. 70022-538X/10/$12.00 doi:10.1128/JVI.02651-09Copyright © 2010, American Society for Microbiology. All Rights Reserved.

Virological Synapse-Mediated Spread of Human ImmunodeficiencyVirus Type 1 between T Cells Is Sensitive to Entry Inhibition�†

Nicola Martin,1 Sonja Welsch,2,4 Clare Jolly,3 John A. G. Briggs,4David Vaux,1 and Quentin J. Sattentau1*

The Sir William Dunn School of Pathology, University of Oxford, Oxford OX1 3RE, United Kingdom1; Structural Biology Unit,Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford OX3 7BN, United Kingdom2; Wohl Virion Centre and

MRC/UCL Centre for Medical Molecular Virology, University College London, London W1T 4JF, United Kingdom3;and Structural and Computational Biology Unit, European Molecular Biology Laboratory, D-69117 Heidelberg, Germany4

Received 30 December 2009/Accepted 12 January 2010

Human immunodeficiency virus type 1 (HIV-1) can disseminate between CD4� T cells via diffusion-limitedcell-free viral spread or by directed cell-cell transfer using virally induced structures termed virologicalsynapses. Although T-cell virological synapses have been well characterized, it is unclear whether this mode ofviral spread is susceptible to inhibition by neutralizing antibodies and entry inhibitors. We show here that bothcell-cell and cell-free viral spread are equivalently sensitive to entry inhibition. Fluorescence imaging analysismeasuring virological synapse lifetimes and inhibitor time-of-addition studies implied that inhibitors canaccess preformed virological synapses and interfere with HIV-1 cell-cell infection. This concept was supportedby electron tomography that revealed the T-cell virological synapse to be a relatively permeable structure.Virological synapse-mediated HIV-1 spread is thus efficient but is not an immune or entry inhibitor evasionmechanism, a result that is encouraging for vaccine and drug design.

As with enveloped viruses from several viral families, thehuman immunodeficiency virus type 1 (HIV-1) can dissemi-nate both by fluid-phase diffusion of viral particles and bydirected cell-cell transfer (39). The primary target cell forHIV-1 replication in vivo is the CD4� T-cell (13), which isinfectible by CCR5-tropic (R5) and CXCR4-tropic (X4) viralvariants (29). R5 HIV-1 is the major transmitted viral pheno-type and dominates the global pandemic, whereas X4 virus isfound later in infection in ca. 50% of infected individuals, andits presence indicates a poor disease progression prognosis(23). Cell-cell HIV-1 transfer between T cells is more efficientthan diffusion-limited spread (8, 16, 32, 38), although recentestimates for the differential range from approximately 1 (42)to 4 (6) orders of magnitude. Two structures have been pro-posed to support contact-mediated intercellular movement ofHIV-1 between T cells: membrane nanotubes (33, 43) andmacromolecular adhesive contacts termed virological synapses(VS) (15, 17, 33). VS appear to be the dominant structureinvolved in T-cell–T-cell spread (33), and both X4 (17) and R5HIV-1 (6, 15, 42) can spread between T cells via this mecha-nism.

VS assembly and function are dependent on HIV-1 enve-lope glycoprotein (Env) engaging its primary cellular receptorCD4 (2, 6, 17). This interaction recruits more CD4 and core-ceptor to the site of cell-cell contact in an actin-dependentmanner (17). Adhesion molecules cluster at the intercellularjunction and are thought to stabilize the VS (18). In parallel,

viral Env and Gag are recruited to the interface by a microtu-bule-dependent mechanism (19), where polarized viral bud-ding may release virions into the synaptic space across whichthe target cell is infected (17). The precise mechanism bywhich HIV-1 subsequently enters the target T-cell cyto-plasm remains unclear: by fusion directly at the plasmamembrane, fusion from within an endosomal compartment,or both (4, 6, 15, 25, 34).

Viruses from diverse families including herpesviruses (9),poxviruses (22) and hepatitis C virus (44) evade neutralizingantibody attack by direct cell-cell spread, since the tight junc-tions across which the these viruses move are antibody imper-meable. It has been speculated that transfer of HIV-1 acrossVS may promote evasion from immune or therapeutic inter-vention with the inference that the junctions formed inretroviral VS may be nonpermissive to antibody entry (39).However, available evidence regarding whether neutralizingantibodies (NAb) and other entry inhibitors can inhibit HIV-1cell-cell spread is inconsistent (25). An early analysis suggestedthat HIV-1 T-cell–T-cell spread is relatively resistant to neu-tralizing monoclonal antibodies (NMAb) (12). A later studyagreed with this conclusion by demonstrating a lack of permis-sivity of HIV-1 T-cell–T-cell spread, measured by transfer ofviral Gag, to interference with viral fusion using a gp41-specificNMAb and a peptidic fusion inhibitor (6). In contrast, anotheranalysis reported that anti-gp41-specific NMAb interfered effec-tively with HIV-1 spread between T cells (26). Inhibitors of theHIV-1 surface glycoprotein (gp120)-CD4 or gp120-CXCR4 in-teraction reduced X4 HIV-1 VS assembly and viral transfer ifapplied prior to mixing of infected and receptor-expressing targetcells (17, 19), but the effect of these inhibitors has not been testedon preformed VS. Thus, the field is currently unclear on whetherdirect T-cell–T-cell infectious HIV-1 spread is susceptible or notto antibody and entry inhibitor-mediated disruption of VS assem-

* Corresponding author. Mailing address: Department of Pathology,The University of Oxford, South Parks Road, Oxford OX1 3R, UnitedKingdom. Phone: 44 1865 275511. Fax: 44 1865 275515. E-mail:[email protected].

† Supplemental material for this article may be found at http://jvi.asm.org/.

� Published ahead of print on 20 January 2010.

3516

on February 25, 2018 by guest

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

bly, and the related question, whether the VS is permeable to viralentry inhibitors, including NAb. Addressing these questions is ofcentral importance to understanding HIV-1 pathogenesis andinforming future drug and vaccine design.

Since estimates reported in the literature of the relativeefficiency of direct HIV-1 T-cell–T-cell spread compared tocell-free spread vary by approximately 3 orders of magnitude(6, 38, 42), and the evidence for the activity of viral entryinhibitors on cell-cell spread is conflicting, we set out to quan-tify the efficiency of infection across the T-cell VS and analyzethe susceptibility of this structure to NAb and viral entry in-hibitors. Assays reporting on events proximal to productiveinfection show that the R5 HIV-1 T-cell VS is approximately 1order of magnitude more efficient than cell-free virus infection,and imaging analyses reveal that the VS assembled by HIV-1 ismost likely permeable to inhibitors both during, and subse-quent to, VS assembly. Thus, we conclude that the T-cell VSdoes not provide a privileged environment allowing HIV-1escape from entry inhibition.

MATERIALS AND METHODS

Cell culture and virus infection. The cell lines Jurkat CE6.1, Jurkat.Tat.CCR5, A201, A301.R5, and ACH-2 were obtained from the National Insti-tute for Biological Standards and Controls (NIBSC) Centre for AIDS Reagents(CFAR), Potters Bar, United Kingdom. A301.R5 cells were derived by retroviraltransduction and cultured as previously described (7). Cells were cultured incomplete medium (CM) composed of RPMI (Gibco/Invitrogen), 10% fetal calfserum (FCS), 100 U of penicillin/ml, and 100 �g of streptomycin (PAA, Yeovil,United Kingdom)/ml. Unless otherwise stated, cells were cultured at 37°C in 5%CO2. Jurkat.Tat.CCR5 and A301.R5 were maintained on CM supplementedwith 1 mg of G418 (Sigma, Poole, United Kingdom)/ml to maintain expressionof CCR5. Peripheral blood mononuclear cells (PBMC) were isolated from wholeblood or buffy coats (from the National Blood Transfusion Service, Bristol,United Kingdom) by Ficoll-Hypaque (Sigma-Aldrich) gradient centrifugation. AMACS CD4 T-Cell Kit-II (Miltenyi UK, Inc.) was used to negatively isolateCD4� T cells that were routinely �99% pure. For experiments with the R5HIV-1BaL isolate, primary CD4� T cells were activated with CM supplementedwith 5 �g of PHA-L (Sigma)/ml and 10 U of interleukin-2 (IL-2; NIBSC)/ml for3 days and then CM with IL-2 alone until use on day 7 of culture. For experi-ments with the X4 HIV-1IIIB isolate, primary CD4� T cells were purified fromPBMC and used immediately. Cells (107) were infected in a volume of 1 ml withIIIB or BaL (NISBC) virus-containing supernatant overnight at a multiplicity ofinfection (MOI) of �0.01 and then cultured in 50 ml of CM and typically usedon days 7 to 9 of infection. At this time point the surface expression of Env washighest, as detected by anti-gp120 antibody 2G12 (NIBSC) and anti-humanIgG-phycoerythrin conjugate (Jackson Laboratories), and analyzed by flow cy-tometry on a FACSCalibur using Flow-Jo software (Becton Dickinson UK).

Inhibitors. The following NMAbs with relatively broad neutralizing activitywere used: 2G12 (CFAR, NIBSC), glycan-specific, binds to mannose groups ongp120 (37, 40); IgG1b12 (B12; D. Burton, Scripps Institute), binds the CD4binding surface on gp120 (5); and 2F5 (D. Katinger, Polymun, Inc.) that bindsthe membrane-proximal extracellular region (MPER) of gp41 and inhibits virus-cell fusion (30). Q4120 is a CD4 domain 1-specific MAb that efficiently blocks theCD4-gp120 interaction and viral entry (14), T20 (enfuvirtide) is a broadly activesynthetic peptide inhibitor of gp41-mediated fusion licensed for human use (28),TAK779 (NIH AIDS Reagent Repository) is a small molecule inhibitor of thegp120-CCR5 interaction (1), and PRO 2000 (Indevus, Inc.) is a polyanioniccandidate topical microbicide that interferes with viral entry, at least in part bybinding to elements of the coreceptor binding surface (36).

Cell-cell and cell-free infection assays. Three different methods were used toseparate infected and target cells to measure cell-free infection, which was thencompared to samples where the cells were directly cultured together to allowcell-cell viral transfer. In each case, 5 � 105 infected cells (JktBaL) and 5 � 105

target cells (A301.R5) were used, and A201 (CD4� CCR5�) were used asnegative control target cells. For infection of target cells by cell-free superna-tants, the JktBaL cells were cultured alone in 150 �l of CM for the 6 or 12 h toproduce virus. Cells were centrifuged at 2,000 � g for 5 min, and the supernatantwas carefully aspirated and immediately added to a culture of 5 � 105 target cells

in 150 �l of CM. Target cells were further cultured to achieve a total incubationtime of 12 h and then centrifuged and harvested for DNA extraction. Forinfection of target cells with or without transwells, cells were either cultureddirectly together or separated by a 3-�m 24-well transwells, allowing full diffusionof virus but not migration of cells [data not shown]). A total of 5 � 105 infectedcells were suspended in 50 �l of CM and placed in the upper well, and 5 � 105

target cells were suspended in 250 �l in the lower well. Target cells wereharvested by removing transwells and lysed and processed for DNA extraction.For the agitated culture method, we used an adaptation of a procedure describedpreviously (42). Briefly, 5 � 105 of each cell type were resuspended in 1 ml of CMin a six-well plate. Paired samples were either left static or rocked at 70 rpm in5% CO2. At the relevant time point, the culture was harvested and processed forDNA extraction for quantitative PCR (qPCR).

Cell-free and cell-cell infection inhibition assays. To measure inhibition ofcell-free or cell-cell infection, 5 � 105 A301.R5 target cells were either culturedwith an equal number of JktBaL cells or 100 ng of PBMC-grown cell-free BaLvirus. At 0 h the inhibitors were titrated onto the cells, starting at 100 �g/ml ina fivefold dilution series for all inhibitors except TAK, which was started at 50nM. Cells were incubated for 12 h (cell-cell) or 24 h (cell-free), corresponding toa single cycle of viral infection. For B12, inhibition was also measured when theNMAb was added at 10 �g/ml at either 1 or 3 h after mixing. The antibody wasdiluted in 7.5 �l of CM and carefully added to existing cocultures of 5 � 105

JktBaL and 5 � 105 A301.R5 target cells to avoid disturbing preformed conju-gates.

Quantitative real-time PCR for HIV-1 polymerase DNA detection. Sampleswere prepared by mixing 5 � 105 infected and 5 � 105 target A301.R5 or controltarget A201.R5 cells in CM and were incubated at for the required time with orwithout the addition of inhibitors. Samples were pelleted, supernatant was aspi-rated, and pellets were frozen at �80°C for later extraction with Qiagen DNeasykit. Primers and probe to detect HIV-1 DNA were designed by using Primer-Express3 against a pol sequence common to BaL and HXB2. The forward primerTGGGTTATGAACTCCATCCTGAT and the reverse primer, TGTCATTGACAGTCCAGCTGTCT were used. The probe had the sequence TTCTGGCAGCACTATAGGCTGTACTGTCCATT and was labeled with the fluorophoreFAM and the quencher TAMRA. For detection of the �-globin cellular refer-ence gene, the forward primer AACTGGGCATGTGGAGACAGA, the reverseprimer CTAAGGGTGGGAAAATAGACCAATAG, and the probe TCTTGGGTTTCTGATAGGCACTGACTCTCTCTG were used. The probe was labeledwith VIC and the quencher TAMRA. PCR was performed in a 25-�l totalvolume, using 1� Eurogentec Mastermix (Eurogentec, Ltd., Southampton,United Kingdom) 0.3 �M concentrations of the primers, 0.1 �M probe, and 2.5�l of sample DNA. A standard curve was created by titrating DNA from ACH-2cells. Reactions were performed in single-plex using an ABI 7500 (AppliedBiosystems, Warrington, United Kingdom) standard run. For all qPCR measure-ments, the results of triplicate single-plex reactions detecting pol and �-globinwere first averaged to give a mean for each, and then the pol values were dividedby the �-globin values to give a relative abundance. For experiments reported inFig. 1A and B, this was further normalized by dividing the value at each timepoint by the relevant 0-h result to adjust for the starting level of infection and toenable comparison between cells with differing efficiency of HIV-1 infection. Forthe experiments whose results are reported in Fig. 1C and D, where the targetcell cultures for cell-free infection did not contain any infected cells, the 0-h valueis shown in the graph. For all other PCR data presented where target andinfected cells were harvested together, the result from a matched sample usingthe nonpermissive A201 cells as targets was subtracted from the test value ateach time point to account for the signal arising from the infected cells. Tocalculate a percent inhibition of HIV-1 infection, the relative abundance of polwas normalized against A201 signal as described above. This value was sub-tracted from the no-treatment control value at the relevant time point, and theresult divided by the no-treatment control value to give percentage inhibition.

Laser scanning confocal microscopy (LSCM). Target A301.R5 cells wereprelabeled with the nonblocking, CD4 domain 4-specific MAb L120 (14, 17),mixed 1:1 with infected cells (JktBaL or JktIIIB), incubated on poly-L-lysine-coated coverslips for the required time, and then fixed at room temperature with4% paraformaldehyde (PFA). Samples were quenched with 50 mM NH4Cl andpermeabilized with phosphate-buffered saline containing 0.1% Triton-X and 5%FCS. Env was labeled with 10 �g of 2G12 (NIBSC)/ml and Gag with 1:1,000polyclonal rabbit anti-p24/anti-p17 (NIBSC). Samples were counterstained withanti-mouse-Alexa 488 (Invitrogen UK), anti-human-TRITC, and anti-rabbit-Cy5(Jackson-Immunotech, United Kingdom) and mounted on glass slides usingProLong Gold (Invitrogen). Images were acquired on a Zeiss Pascal Axiovert200M by using sequential capture and processed using Adobe Photoshop CS2.Gag polarization and transfer between cells were quantified in 10 randomly

VOL. 84, 2010 INHIBITION OF HIV-1 CELL-CELL SPREAD 3517


http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

acquired low-power fields, each containing an average of approximately 20 con-jugates. Conjugate interfaces were defined as contacts between an infected(Env� and/or Gag�) and a target (CD4�) cell. Polarization of Gag, Env, or CD4was defined as signal enrichment in the membrane contact zone compared to thenoncontact regions quantified using Metamorph software. VS were defined bycolocalization of all three markers at the intercellular interface, and percentageswere calculated by dividing the number of colocalization events by the number ofinterfaces. Gag transfer was defined as Gag signal within the target cell in theabsence of colocalized Env signal. The total number of these events/field wasdivided either by the number of VS or by the total number of conjugate inter-faces.

Time-lapse LSCM analysis of conjugate lifetimes. To measure the duration ofinteraction times between HIV-1-infected and uninfected CD4� target cells,JktBaL cells were labeled with CellTracker Orange CMRA (Invitrogen) anduninfected primary CD4� T cells were labeled with carboxyfluorescein diacetatesuccinimidyl ester (CFSE; Invitrogen). A total of 104 JktBaL and 104 primaryCD4� T cells were mixed in 50 �l of CM and placed in one well of an IbidiIbiTreat 24-well angiogenesis slide (Ibidi GmbH, Munich, Germany), and theplastic cover was replaced. The slide was placed within a heated stage andsupplemented with humidified 5% CO2. Using a Zeiss Pascal LSM4 with se-quential channel capture, images were recorded every 1 min up to 5 h. Toquantify infected cell-target cell interaction times, each frame of the time-lapseseries was processed in Metamorph to binarize the cell color labels into red(infected) and green (target), the cell profile was dilated to create definedinteraction zones (yellow) at cell-cell contact regions, and interaction sites wereisolated for quantification. Interaction sites for each x-y frame of the movie werestacked into a three-dimensional box with time along the z axis, allowing eachinteraction to be tracked over time as an individual object.

Electron microscopy and tomography. Freshly purified primary CD4� T cellswere mixed 1:1 with JktIIIB cells, or 3-day PHA/IL-2-activated primary CD4� Tcells were mixed with JktBaL cells at 106 cells/ml in RPMI–10% FCS for 3 h at37°C in 5% CO2, followed by gentle pelleting, aspiration of the supernatant, andfixation in 4% PFA for 10 min at 37°C and 8% PFA for 50 min at roomtemperature. Samples were left in 2.5% glutaraldehyde in cacodylate buffer (100mM sodium cacodylate, 50 mM KCl, 2.5 mM CaCl2 [pH 7.2]) overnight, post-fixed for 1 h on ice with 1% osmium tetroxide, washed, stained overnight at 4°Cwith 0.5% uranyl acetate, and dehydrated in a graded ethanol series at roomtemperature. Samples were embedded in epoxy resin (Roth, Karlsruhe, Ger-many) according to the manufacturer’s instructions. Serial sections (300 nmthick) sections were cut, and dual-axis digital image series were recorded on anFEI Technai TF30 microscope at 300 kV with FEI Eagle 4k charge-coupleddevice camera (pixel size after binning to 2k 1.99 or 2.53 nm at the specimenlevel) over a �60° to �60° tilt range (increment 1°) and defocus at �0.2 �m,using the SerialEM software package (27). Tomogram reconstruction, joining ofserial tomograms, manual three-dimensional image segmentation and calcula-tion of membrane model point coordinates used IMOD software v3.12.20 (20).Nearest approach points between the two membrane models in each tomogramwere calculated and displayed as colored patches using MATLAB (Mathworks,Natick, MA) and UCSF Chimera (31), respectively.

Statistical analyses. Unless otherwise stated, quantitative data represent themean values of replicate independent experiments each carried out in triplicate,and error bars represent the standard errors of the mean (SEM). Significancetesting throughout was performed using a two-tailed Student t test, with theBonferroni correction for multiple tests where appropriate.

FIG. 1. R5 HIV-1 spreads more efficiently cell-cell than by cell-freevirion diffusion. (A) HIV-1BaL transfer between infected and unin-fected T-cell lines. JktBaL was mixed with A301.R5 target cells and, atthe time points indicated, the samples were lysed and analyzed forHIV-1 pol by qPCR and then normalized against cellular �-globin. Thebackground signal obtained from nonpermissive A201 cells was sub-tracted from all values before calculation of percent inhibition com-pared to no-treatment or no-virus controls. (B) HIV-1 transfer fromprimary CD4BaL cells to autologous activated primary CD4� T cells,assayed by qPCR and presented as in panel A over 24 h. The datarepresent means of eight independent experiments, each carried out intriplicate. (C) JktBaL cells were mixed directly with A301.R5 targetcells in a transwell chamber without membrane (�) or were separatedby the transwell membrane (f). At the times shown cultures were lysedand processed for qPCR. The data represent the mean pol copy num-ber relative to �-globin from three independent experiments, eachcarried out in triplicate. (D) Supernatant was harvested from the lower(containing A301.R5 target cells) chamber of a transwell without (�)or with (f) a transwell membrane to separate the top well (containing

JktBal cells) and assayed for p24 Gag by ELISA. Triplicate wells weresampled from one representative experiment and the bars representthe mean � 1 standard deviation (SD). (E) JktBaL (white bars) orcell-free supernatants harvested from JktBaL cells over a 12-h culture(f) were mixed with A301.R5 target cells at T � 0. Samples wereprocessed as in panel A, and the data represent the mean pol copynumber relative to �-globin from three independent experiments eachcarried out in triplicate. (F) JktBaL cells were mixed with A301.R5target cells at T � 0, after which the cultures were either left static (�)or were agitated to prevent stable cell-cell contacts from forming (f).Samples were processed as in panel A, and the data represent themean pol copy number relative to �-globin from three independentexperiments, each carried out in triplicate. Error bars represent theSEM. *, P 0.05; **, P 0.01; ***, P 0.001.

3518 MARTIN ET AL. J. VIROL.


http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

RESULTS

Efficiency of direct T-cell–T-cell spread of R5 HIV-1. SinceR5 HIV-1 is the virus type that is principally transmitted be-tween individuals, its study is central to understanding viralinfection and spread and inhibition of these processes. Wetherefore adapted a previously established qPCR-based assayfor detection of de novo DNA viral transcripts in X4 virus-infected cells (19) for use with R5 viral DNA. The pol regionwas chosen as a template for primers and probe since this is alate reverse transcript and therefore relevant to infectionevents proximal to proviral integration and productive replica-tion. Values normalized to �-globin were expressed as copies ofHIV-1 DNA/cell, and assay specificity was confirmed by cocul-turing CD4� CCR5� Jurkat.Tat.R5 cells infected with HIV-1BaL (JktBaL) with either the permissive CD4� CCR5� T-cellline A301.R5, or the nonpermissive CD4� CCR5� line A201.As previously observed with X4 HIV-1 (18, 19), de novo polDNA production was not observed in A2.01 target cells (datanot shown), whereas A3.01.R5 target cells demonstrated asignificant (P 0.05) increase of �5-fold over background inpol DNA over 12 h (Fig. 1A). Because model systems usingimmortalized T-cell lines may not accurately represent inter-actions taking place between primary CD4� T cells, we mea-sured relative R5 HIV-1 pol production after transfer betweeninfected and uninfected primary CD4� T cells. Activated pu-rified primary CD4� T cells infected to levels �50% as deter-mined by Gag and Env staining and flow cytometric analysis(data not shown) were mixed with uninfected autologousCD4� T cells for the times shown, and the samples wereprocessed for qPCR. We observed significant levels of in-creased pol in mixtures of infected and uninfected primaryCD4� T cells (Fig. 1B) that were similar in magnitude to virusspread between the cell lines, but with the slower kinetics of polsynthesis. This confirmed that the use of T-cell lines appearsqualitatively justified for the study of cell-cell spread of R5HIV-1, with the caveat that it is more rapid than the primarycell system. The HIV-1BaL-infected T-cell line JktBaL was sub-sequently used in all experiments hereafter for reasons of re-producibility, and A301.R5 cells were used as targets unlessotherwise stated.

To quantify differences between cell-free and cell-cell R5HIV-1 transmission, we used three different assay formats: (i)use of a virus-permeable transwell to separate JktBaL fromtarget cells; (ii) culture of A3.01/R5 target cells with superna-tants from JktBaL or direct A3.01/R5 mixing with JktBaL; and(iii) an adaptation of the approach adopted by Sourisseau et al.(42), in which cultures were maintained static for cell-cell in-teractions to take place or were agitated to prevent stablecell-cell interactions. We confirmed that in each assay infectedand target cell viability was maintained at an equivalent levelover the time course of the experiment (data not shown).Figure 1C shows that by 12 h after JktBal-target cell mixingthere was a robust pol signal detectable that was �13-foldgreater than the cell-free signal (P 0.01). Free diffusion ofvirus across the transwell membrane was determined by sam-pling supernatant for p24 Gag enzyme-linked immunosorbentassay (ELISA): there was no significant difference in p24 levelsabove or below the membrane (Fig. 1D). The signal obtainedwhen JktBal-target cells were mixed was 6.5-fold greater than

the signal obtained with viral supernatant (Fig. 1E, P 0.01)and 5.4-fold greater when cells were maintained in a static, asopposed to an agitated, culture (see Fig. 7F, P 0.01). Takentogether, the mean increase in cell-cell compared to cell-freeinfection at 12 h corresponds to (8.3 3.3)-fold. Although wedo not know which assay most precisely mimics the in vivosituation relating to cell-cell compared to cell-free viral spread,we assume that the mean value of approximately 10-fold isbroadly representative of relative in vitro efficiency of HIV-1spread between T cells.

Cell-free and cell-cell spread of R5 HIV-1 are equally sus-ceptible to entry inhibition. Having established that three dif-ferent cell-cell assays for R5 virus spread gave broadly similarresults and having compared the efficiency of cell-free andcell-cell spread quantitatively, we next analyzed whether thesemodes of infection are similarly susceptible to interferencewith viral entry. We chose the supernatant-based cell-free as-say alongside the cell-cell infectivity assay (as shown in Fig. 1E)for these experiments since this was the most technically fea-sible for this type of inhibition experiment and gave an inter-mediate result (6.5-fold) in terms of relative viral transfer ef-ficiency. We selected a panel of HIV-1 entry inhibitors basedupon their modes of action (receptor blocking or fusion inhi-bition) their chemical type (neutralizing antibody, peptide,small molecule inhibitor, or polyanion) and their relative mo-lecular mass (0.5 to 150 kDa). Inhibitors were titrated intoeach system at time of cell-free infectious supernatant or in-fected cell mixing with target cells. De novo-synthesized viralDNA was measured by qPCR within a time-frame allowingrobust pol detection and corresponding to a single round ofviral replication in both systems, approximating 12 h in thecell-cell system and 24 h in the cell-free system. Titrationcurves from combined experiments are shown for each inhib-itor tested in Fig. 2A, and the mean 50% inhibitory concen-tration (IC50) data are summarized in Fig. 2B. All agentsshowed increasing inhibition of both cell-cell and cell-free in-fection with increasing concentration, and all achieved ca.100% inhibition at the higher doses in the range. Mean ID50

values varied across 4 orders of magnitude, representing theirrelative potency against R5 HIV-1. Considerable biologicalvariation was observed between experiments, particularly withthe CCR5 antagonist TAK779, for reasons that are currentlyunclear. Two of the inhibitors that interfere with gp120-CD4binding (B12 and Q4120) showed a clear trend at low concen-trations toward less effective inhibition of cell-cell viral spreadcompared to cell-free spread (Fig. 2A). This might reflect amodest selective reduction in inhibition of cell-cell spread as aresult of reduced inhibitor access to the VS, increased velocityof gp120 engagement of CD4 on the target cell, or both. Nev-ertheless, no overall significant difference in inhibitor efficacyas measured at the ID50 was observed between the two sys-tems, whether inhibitors of low (TAK779, �0.5 kDa [1]) orhigh (NAb, �150 kDa) molecular mass were tested. This im-plies that the inhibitors were either acting rapidly to preventVS assembly and cell-cell virus transmission, or that thesemolecules enter assembled VS to block viral infection, or both.

Life span of HIV-1 R5 T-cell VS. To investigate the possi-bility that inhibitors might interfere with HIV-1 spread acrossexisting VS, we first established the longevity of R5 VS in Tcells. For this we assumed that long-lived conjugates formed



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

3520


http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

between HIV-1-infected and uninfected, receptor-expressingCD4� T cells represented VS, a proportion of which would befunctional in cell-cell transfer of HIV-1 infection. Differentiallylabeled JktBaL cells or control uninfected Jurkat cells weremixed with uninfected primary CD4� T cells, and their inter-actions were imaged using time-lapse LSCM. A frame from arepresentative movie (see Movie S1 in the supplemental ma-terial) is shown in Fig. 3A, in which multiple conjugates be-tween infected (red) and uninfected target cells (green) areidentified by arrows. To quantify infected cell-target cell inter-action times, each frame of the time-lapse series was processedto create binary versions of the two colors and to dilate the cell

profile to create defined interaction zones (yellow) at cell-cellcontact regions (Fig. 3B). Interaction sites were then isolatedfor quantification (Fig. 3C). Interaction sites for each x-y frameof the movie were stacked into a 3-dimensional box with timealong the z axis, allowing each interaction to be tracked overtime as an individual object. Lifetimes of control interactionsobtained from uninfected Jurkat and primary CD4� T cellslabeled and processed in the same manner as the test sampleswere compared to the test, and the data are quantified in Fig.3D and E. Control and test sample conjugate life spans weresimilarly represented at times of up to 10 min and are consis-tent with previously described intravitally measured life spans

FIG. 2. Cell-free and cell-cell infections are equally susceptible to entry inhibition. (A) A301.R5 target cells were either mixed 1:1 with JktBaLcells or with 12-h cell-free supernatants derived from JktBaL in the presence of titrations of the inhibitors shown. Cells were harvested after 12 hand processed for qPCR of pol and �-globin. Reduction in the pol PCR signal relative to the �-globin signal was expressed for each dilution of eachinhibitor. The background signal obtained from nonpermissive A201 cells was subtracted from all of the values before calculation of the percentinhibition compared to no-treatment or no-virus controls. Solid lines represent cell-free virus inhibition, and dashed lines represent cell-cell virusinhibition. (B) Percentage inhibition data from three independent experiments each carried out in triplicate were used to calculate IC50s. �,Cell-free infection inhibition; f, cell-cell infection inhibition. Error bars represent the SEM.

FIG. 3. Life span of the R5 HIV-1 VS. (A) JktBaL cells prestained with CTO (red) and activated primary CD4� T cells prestained with CFSE(green) were mixed in a 1:1 ratio and seeded onto MatTek poly-D-lysine-coated glass-bottom tissue culture dishes. Cells were imaged by time-lapseLSCM, with one frame acquired every 30 s for 250 min. A representative frame is shown with red, green, and differential interference contrastmerged, with cell-cell interactions indicated by yellow arrows. (B) Same two-color image as in panel A after setting the threshold, binarizing thegreen and red images, and dilating the individual cell profiles as described in Materials and Methods. Cell-cell interaction sites are visible as yellowcrescents. (C) The interactions are isolated in x-y two-dimensional frames as discrete white regions. Individual x-y frames were stacked togetherinto a three-dimensional box, allowing tracking of the individual events over time. (D) The test and control series were processed as described inMaterials and Methods to give the duration of each cell contact, which were grouped into the temporal categories. (E) The events falling into the�10-min category for the test movie are depicted as individual points, with the red line representing the mean duration in min.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

of normal homotypic T-cell interactions (35). Since �100% ofthe JktBaL cells were infected, this implies that many infected-target cell interactions are similarly transient to uninfectedcell-cell interactions: whether these rapid interactions result inviral transfer is unknown. However, none of the control inter-actions exceeded 10 min, demonstrating that the formation oflong-lived T-cell–T-cell contacts is HIV-1-driven. The meancontact time in the test samples was 62 min, and a smallnumber were stable to the end of the assay at 300 min. Sincethese sustained (�10-min) interactions representing �21% ofall events approximate the percentage of VS found withinT-cell conjugates in other studies with X4 virus (17), we hy-pothesized that these were likely to represent VS. However,since X4 and R5 HIV-1 VS kinetics of assembly and frequency

may vary, we investigated R5 virus VS dynamics in the currentsystem using LSCM of JktBaL-primary CD4� T-cell conjugates.The percentage of CD4, Gag, and Env capping at differenttime points was quantified from randomly selected low-powerfields as previously described (17). All three markers indepen-dently showed increased capping from 1 to 3 h after cell mix-ing, plateauing between 3 and 6 h (Fig. 4A). Similarly, cocap-ping of all three markers at a conjugate interface (ourdefinition of a VS) increased significantly between 1 and 3 hbut not between 3 and 6 h, reaching a maximum value of�30% at 6 h. Gag transfer across the VS measured by distinctlocalization of Gag within the target cell revealed increasingtransfer over the 6-h period. When this was expressed as per-centage of total VS, ca. 20% VS showed Gag transfer at 1 h,

FIG. 5. Time of inhibitor addition effects on VS formation. (A) Target A301.R5 cells prelabeled with the nonblocking CD4 MAb L120 weremixed 1:1 with JktBaL and inhibitor was either added simultaneously (white bars) or after 1 h of cell coculture (black bars). The hatched barsrepresent VS assembly in the absence of inhibitor (Non-Treated Control, NTC). Cells were cultured for a total of 3 h before fixing, permeabilizingand labeling for Gag and Env. Randomly selected conjugate interfaces were analyzed for colocalization of the three markers defined as VS, asdescribed for Fig. 4. The data are means of values from three independent experiments each carried out in triplicate, and error bars represent SEM.(B) As for (A) except that the infected cells were JktIIIB. P 0.05.

FIG. 4. Assembly dynamics of the R5 HIV-1 VS. (A) A301.R5 target cells prelabeled with the nonblocking CD4 MAb L120 were mixed withJktBaL cells for the times shown before fixing, permeabilizing, and labeling with the appropriate secondary reagent for CD4 (u) or with theappropriate primary and secondary reagents for Gag (s) or Env (f). Labeling and colocalization of all three markers in the same samples wasalso carried out (�). The data represent the combined percentage values from multiple conjugates counted in randomly selected low-power fields,and error bars represent � 1 SD. *, P 0.05. (B) Same as in panel A but the percentage of VS in which Gag was observed independently of Envin the target cell was divided by the number of VS (�) or by the total number of conjugate interfaces (s).



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

rising to �60 and �80% at 3 and 6 h, respectively (Fig. 4B).Expression of these values as a percentage of all conjugateinterfaces showed accumulation of target cell-associated Gagto a plateau of ca. 20% between 3 and 6 h (Fig. 4B), a findingconsistent with the percentage of long-lived conjugates seen inFig. 3E. We therefore hypothesize that the majority of thelong-lived conjugates forming between HIV-1-infected andCD4� target cells are likely to contain a VS functional for Gagtransfer and that the average functional VS lifetime is probably�1 h.

Inhibition of HIV-1 cell-cell spread in preformed VS. Withthis knowledge we next addressed the question of whetherinhibitors added after conjugate formation interfere with R5HIV-1-driven VS assembly. Since this concept has not beeninterrogated for X4 HIV-1 spread across VS, we investigatedthis in parallel, basing time of addition on the data reportedpreviously (17) that demonstrate X4 HIV-1 VS assemblyreaching a plateau in ca. 30% of conjugates by 3 h after in-fected-target cell mixing. JktBaL- or JktIIIB-A3.01.R5 conju-gates were prepared, and inhibitors at a single inhibitory con-centration (10 �g/ml for antibodies, 50 �M for PRO 2000, and1 �M for TAK 779 and AMD3100, which had previously beenshown to be close to 100% inhibitory when added at 0 h [Fig.2A]), were added either at the time of cell mixing or at 1 h aftercell mixing when ca. 30% of VS had formed (Fig. 4A). Con-jugates were then fixed at 3 h when VS frequency was at aplateau (Fig. 4). Multiple fields were scored for the percentageof conjugates forming VS, and pooled data are shown in Fig. 5.Approximately 30% of JktBaL-A3.01.R5 conjugates formed VSover the 3 h time course in these experiments (Fig. 5A), afinding consistent with the data reported in Fig. 4A, and a verysimilar number was also observed for JktIIIB-A3.01.R5 conju-gates (Fig. 5B). When added at T � 0, R5 HIV-1-mediated VSassembly was most strongly inhibited by blockers of the CD4-gp120 interaction such as B12 (�60% inhibition) and Q4120(�80% inhibition), as previously observed for the X4 VS (17).In contrast, antagonists of the gp120-CCR5 interaction(TAK779 and PRO 2000) were more weakly inhibitory. Whenadded after 1 h of coculture, VS inhibition was significantlyreduced for B12 and Q4120, whereas no significant increase inVS number was observed with the other inhibitors. Thus, onceformed, the R5 VS appears moderately resistant to disassem-bly by HIV-1 receptor blockers. In contrast, the effect of CD4-gp120 or CXCR4-gp120 antagonists on the X4 HIV-1 VS wasmore pronounced. Whether added at 0 or 1 h, all inhibitorswere �75 to 100% effective at inhibiting VS assembly, suggest-ing that the preformed X4 VS may be susceptible to disassem-bly by antagonists of gp120-receptor interactions. Taken to-gether, these data imply that R5 VS assembly is less easilyinhibited than the X4 VS by blockers of gp120-receptor inter-actions, whether applied before or during VS formation. Therelative resistance of the R5 VS structure to inhibition mayrepresent an inability of inhibitors to access a preformed VSor, alternatively, may demonstrate that preformed VS struc-ture may be maintained independent of function in terms ofHIV-1 cell-cell spread. To investigate these possibilities, wechose a single high-molecular-mass (�150-kDa) inhibitor, theNMAb B12, for further analysis. Since B12 prevents most VSassembly when introduced at the time of infected-uninfectedcell mixing, we hypothesized that the presence of B12 within a

FIG. 6. Effects of NMAb B12 on HIV-1 infection across VS.(A) Target A301.R5 cells prelabeled with the nonblocking CD4-specific MAb L120 were mixed with JktBaL cells and cocultured onpoly-L-lysine-coated coverslips. After 1 h of coculture, B12 wasgently added to yield a final concentration of 10 �g/ml. The cellswere cocultured for a further 2 h and then fixed, permeabilized, andlabeled for B12 with anti-human IgG-TRITC and for Gag withanti-p24 antiserum, and Gag and CD4 were labeled with the ap-propriate secondary detection reagents. Coverslips were mountedand analyzed by LSCM. The asterisk labels the infected cell.(B) Target A301.R5 cells were mixed 1:1 with JktBal cells in thepresence of 10 �g of B12/ml (T � 0), or the same concentration ofB12 was added 1 h or 3 h after cell mixing. Cells were then culturedfor a total of 12 h prior to lysis and qPCR for pol and �-globinproducts. The results are expressed as the percent inhibition ofrelative pol synthesis compared to no inhibitor or mixtures of JktBaLcells with the nonpermissive A201 target cell control. Each barrepresents the mean data from three independent experiments car-ried out in triplicate, and error bars represent the SEM.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

FIG. 7. Electron tomographic reconstruction of the HIV-1 T-cell VS. (A) A single 2-nm-thick digital slice from a tomographic reconstructionof a conjugate between a primary CD4� T cell (left) and a JktIIIB cell (right) cocultured at a 1:1 ratio for 3 h. (B) Single higher-magnification5-nm-thick digital slice from a tomographic reconstruction of a conjugate between a primary CD4� T-cell (left) and a JktBaL cell (right). Arrowsindicate viral budding structures on the infected cell; the white asterisk indicates the nucleus of an infected cell. (C) Another single digital slice



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

VS would most likely signify incorporation of NMAb into apreexisting VS. JktBaL and A3.01 cells were cocultured for 1 h,which results in at least 30% of total VS formation (Fig. 4A),prior to the addition of B12 to the culture, and the cultureswere maintained for a further 2 h prior to fixing and labelingfor LSCM. Figure 6A shows a representative conjugate inwhich B12 has robustly labeled the conjugate interface andcolocalizes with CD4 on the target cell and viral Gag in theinfected cell. These data support the notion that inhibitors ofrelatively high molecular mass can enter a VS without disas-sembling the supramolecular structure. To investigate thefunctional effects of inhibitors on preformed R5 HIV-1 VS, weset up a experiment similar to that described for Fig. 6A withthe same concentrations of inhibitors added either at the timeof cell mixing or 1 or 3 h later but lysed the cells after a totalof 12 h of culture and analyzed them for the relative pol signal.The addition of inhibitors at the time of mixing resulted in 70%inhibition of the relative pol signal (Fig. 6), a finding consistentwith the results in Fig. 2. Taken together, the results presentedin Fig. 5 and 6 suggest that inhibitors can either prevent VSassembly if present during initial infected cell-target cell con-tact or if introduced subsequent to VS formation can enterintact VS and inhibit viral transfer into the target cell andsubsequent reverse transcription.

Ultrastructure of the VS. Although the data shown in Fig. 5and 6 are suggestive of a VS structure permeable to inhibitors,this is not unequivocal since some inhibitor may conceivably beincorporated into forming VS without necessarily blocking fur-ther assembly. We therefore used electron tomography toprobe the structure of mature R5 and X4 HIV-1 VS. Conven-tional thin-section electron microscopy yields ultrastructuraldetails of cellular interactions in two dimensions but is unableto resolve the three-dimensional details of a complex intercel-lular interface. We therefore carried out electron tomographyof serial thick sections spanning �1.2 �m total thickness, ofepoxy-embedded conjugates between infected Jurkat cells andprimary CD4� T cells fixed 3 h after cell mixing. The associ-ated single-section movies are Movies S2 and S3 in the sup-plemental material. Figure 7 shows single slices from repre-sentative conjugates formed between JktIIIB (Fig. 7A to C) orJktBaL (Fig. 7D and E) and primary CD4� T target cells. Theinfected cells were identified based on the presence of buddingstructures in HIV-1IIIB (Fig. 7B and C)- and HIV-1BaL (Fig.7E)-infected cells. The plasma membrane interfaces for thetwo VS are clearly defined, with multiple virions and virus-likevesicles present at the interface between the cells but fewpoints of obvious intracellular contact. Three-dimensionalmodels reconstructed from the tomographic analysis of theserial sections are shown in Fig. 7F and G. It is evident that theVS interface for both virus types is complex with multiple

membrane invaginations and projections, and yet the interfacebetween the cells appears loosely structured. A large numberof X4 HIV-1 virions are present in gaps between the two cells,many of which are adherent to the target cell plasma mem-brane, presumably via Env-receptor and/or adhesion interac-tions. Fewer virions can be seen in the R5 HIV-1-infected VS,and most of these are either free or are associated with theinfected cell membrane, possibly reflecting a lower avidity fortarget cell receptors than the X4 virions. The infected-targetcell plasma membranes are relatively distant (�100 nm) overmost of the synapse but are punctuated by limited areas ofclose membrane apposition. Analysis of the distance betweenthe target cell and the nearest infected cell plasma membranewith virions removed to simplify analysis revealed regions thatwere 30 nm or closer (Fig. 7H and I), a finding consistent withintegrin interactions such as those which stabilize the immu-nological synapse (41) and implicated in X4 HIV-1 VS struc-ture and function (18). These discrete adhesive regions weretypically concentrated between interlocking villi surrounded bylarge areas of noninteracting plasma membrane. We have ob-served that the plasma membranes are often dramatically ruf-fled or invaginated (e.g., target cell membrane in Fig. 7D),which might lead to the interpretation, from a single section, ofa large internal virus-containing compartment. However, therenderings in Fig. 7G and I imply that this is not a closedcompartment but open to the external milieu, and many virionscan be seen at the edges of and outside the zone of cellmembrane contact. Despite analysis of multiple tomographicreconstructions, we have yet to observe virus within a closedcompartment within the target T cell. Nevertheless, since thesections that we have analyzed generally span only a propor-tion of the entire VS area, we cannot exclude the possibilitythat some virus is internalized into internal compartmentswithin the target cell. In conclusion, these models reveal aporous intercellular interface containing mature virus particlesthat would allow access of even high-molecular-weight com-pounds to virus without the imposition of obvious steric con-straints.

DISCUSSION

Enveloped viruses from several different families use directpassage between cells to propagate themselves (39), andHIV-1 is no exception. We found here using three differentassays that cell-cell spread is approximately an order of mag-nitude more efficient than cell-free spread, a figure that agreeswith one recent study (42) but disagrees with another (6). Webelieve that the discrepancy comes from the viral endpoint ofcell-cell spread measured: the Sourisseau study assayed forproductive infection by measuring Gag release into the super-

at a different z height within the same tomogram as in panel A. Arrows indicate budding structures. (D) Same as in panel A but an image froma conjugate between a JktBaL cell and a primary CD4� T cell. (E) A higher-magnification image represents a diagonal slice through thethree-dimensional volume for better visualization of the budding structure (arrow) on the surface of the JktBaL cell. (F) Three-dimensional modelof the JktIIIB cell surface (red), the primary CD4� T-cell surface (green), and virions (red spheres), generated from the same tomogram as in panelA. (G) Three-dimensional model of the JktBaL cell surface (red), the primary CD4� T-cell surface (green), and virions (red spheres), generatedfrom the same tomogram as in panel D. (H) Higher-power representation of cell membranes shown in the boxed area of panel F. Yellowmembrane patches indicate areas of close apposition (30 nm) of the two membranes. (I) Higher-power representation of cell membranes shownin the boxed area of panel G; the yellow label is as described in panel H.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

natant, whereas the Chen study quantified Gag transfer intotarget cells. Gag transfer does not necessarily correspond toinfection, particularly since the authors report that much of theGag appears to be endocytosed by the target cell, an entrypathway that may result in a high proportion of noninfectiousevents (2, 4, 6, 15). Like the Sourisseau study, we assayed fora proximal surrogate of productive viral infection, synthesis oflate viral DNA product, which we believe is most likely torepresent an infectious outcome in the target cell. Divergentexperimental conditions may also influence the rate and extentof viral transfer between T cells, including the ratio of infectedto uninfected T cells, the chronicity of the infection in thedonor cells, and the type of cell, or cell line used.

The mechanism of enhanced efficiency of cell-cell spreadcompared to cell-free spread of HIV-1 remains to be formallyelucidated. Studies by others (8, 16, 32, 38) and our datapresented here and previously published (17), however, sug-gest that there may be several factors determining efficiency ofviral dissemination in these systems. (i) The first factor is theproximity of the target and effector cells. Since the infected anduninfected cells are in relatively close and stable appositionduring cell-cell spread, the rate-limiting factor of fluid-phasediffusion is essentially eliminated. (ii) Receptor clustering is anadditional factor. The VS is defined as the polarization of CD4and coreceptor to the site of infected-target cell contact. Thiswould allow de novo budded HIV-1 to engage an optimalnumber of receptors rapidly without the delay implicit in re-ceptor recruitment required for fusion that is imposed uponindividual cell surface-associated virions (21). (iii) Finally, wemust consider the MOIs of target cells. The rate of spread ofHIV-1 in culture is directly related to the MOI (8). Althoughwe have not been able to quantify the number of infectiousvirions traversing a VS, the MOI via VS will be much higherthan that occurring via cell-free spread because of both theincreased total number of particles engaged by the target celland the reduced time-dependent inactivation of viruses under-going cell-cell spread compared to cell-free spread. We intendthat the assays used here reflect as far as possible the in vivosituation regarding viral spread between lymphocytes. One ob-vious difference is that our cultures are �100-fold less densethan the estimated lymphocyte density in vivo (45). Such a highin vivo density of lymphocytes would increase the chances ofboth cell-cell and cell-free infection occurring, however, mak-ing it difficult to model how this might affect the differentialefficiency of viral spread by the two modes.

Our demonstration that a panel of HIV-1 entry inhibitorsinterferes equivalently with cell-cell and cell-free HIV-1 infec-tion when inhibitors are added at the time of mixing target cellswith infected cells or free virus suggests that VS-mediatedspread of HIV-1 between T cells is unlikely to be an antibodyor drug evasion strategy for the virus. In most scenarios in vivo,the inhibitor (antibody or therapeutic entry inhibitor) would bepresent prior to addition of virus to cells and so would not haveto enter a preformed VS in order to interfere with cell-cell viralspread. However, even if inhibitor was added subsequent toencounter between a virus-infected T cell and an uninfected Tcell, our data suggest that such an inhibitor could access thepreformed VS and prevent viral cell-cell spread. These resultsare encouraging for use of prophylactic vaccines designed toelicit neutralizing antibodies and for entry inhibitors applied in

a prophylactic or therapeutic setting. Our results showing thatall entry inhibitors tested target both X4 and R5 HIV-1 cell-cell spread similarly to diffusion-limited spread are inconsistentwith certain of the data obtained in T-cell systems by others(2–4, 6). We consider that the most likely explanation againrelates to systems measuring viral antigen uptake by endocyticpathways that are resistant to viral coreceptor antagonists andfusion inhibitors, but their contribution to productive infectionis unclear. In this respect the same group that reported anabsence of inhibition of VS-mediated Gag transfer by corecep-tor antagonists in one study (6) also reported inhibition by thesame agent in a later study when a better surrogate of viralinfection (transcription of HIV-1 long-terminal-repeat-drivengreen fluorescent protein) was measured (15). Although whenusing electron microscopy or tomography we have not ob-served evidence of endocytic uptake of HIV-1 by primaryCD4� T cells across a VS in the present study or previously(17), our analyses do not exclude this as an outcome of cell-celltransfer of virus. However, we assume that such events may berare in primary CD4� T cells compared to the immortalizedcell lines used predominantly in other analyses (2–4, 6).

Our measurements of uninfected CD4� T-cell–T-cell inter-actions are consistent with previously described intravitallymeasured life spans of homotypic T-cell interactions (35) andconfirm that T cells undergo transient adhesive interactionsunder normal conditions. A proportion of HIV-1-infected Tcells undergo longer interactions, however, and we assume thata percentage of these relates to VS assembly and viral cell-celltransfer. A small minority of events lasted �300 min: thesemay be very long-lived interactions, perhaps resulting fromhigh-level infection of the donor T-cell leading to enhancedcell-cell interactions, and are consistent with observed conju-gate life spans (18 to 32 h) reported earlier (15). Alternatively,these long-lived interactions may represent a small proportionof conjugates undergoing membrane fusion to form syncytia.

Although the electron tomographic analyses carried outhere and thin-section electron microscopy carried out previ-ously (17) reveal the T-cell VS to be loosely structured, othercell types may assemble VS-like structures with distinct func-tional and structural features. We recently described a VSformed between HIV-1-infected macrophages and CD4� Tcells: thin-section electron microscopic analysis revealed abroad interaction surface between the cells that had the ap-pearance of a tight junction but that formed only transiently(11). A study evaluating the ultrastructure of the VS formedbetween human T cell leukemia virus type 1 and permissivetarget T cells also found large surfaces of tightly apposedmembrane with occasional pockets containing virions (24).Thus, the type of VS assembled is likely to be both cell andvirus type specific, and this may influence sensitivity of thismode of viral spread to inhibition. Indeed, in this respect,dendritic cell transfer of HIV-1 to CD4� T cells in trans acrossinfectious synapses has been demonstrated to be resistant to avariety of NMAb (10, 46). Further comparative analyses of VSformed between different cell types, and their sensitivity toNAb and entry inhibitors are therefore warranted.

ACKNOWLEDGMENTS

We thank the electron microscopy core facility at EMBL Heidelbergfor equipment and assistance with electron tomography; Sylvain La-



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

comble for assistance with image processing; and Leo Kong, BeckyRussell, Will Hillson, and Kate Gartlan for proofreading the manu-script.

This study was supported by the Medical Research Council UK, theInternational AIDS Vaccine Initiative Neutralizing Antibody Consor-tium, EUROPRISE, and Fondation Dormeur. S.W. and the electrontomography part of this study were funded by Wellcome Trust grantH5RCYV0 to Stephen Fuller and Deutsche Forschungsgemeinschaftgrant SPP1175 to J.A.G.B. Q.J.S. is a Jenner Institute Fellow.

REFERENCES

1. Baba, M., O. Nishimura, N. Kanzaki, M. Okamoto, H. Sawada, Y. Iizawa, M.Shiraishi, Y. Aramaki, K. Okonogi, Y. Ogawa, K. Meguro, and M. Fujino.1999. A small-molecule, nonpeptide CCR5 antagonist with highly potent andselective anti-HIV-1 activity. Proc. Natl. Acad. Sci. U. S. A. 96:5698–5703.

2. Blanco, J., B. Bosch, M. T. Fernandez-Figueras, J. Barretina, B. Clotet, andJ. A. Este. 2004. High level of coreceptor-independent HIV transfer inducedby contacts between primary CD4 T cells. J. Biol. Chem. 279:51305–51314.

3. Bosch, B., J. Blanco, E. Pauls, I. Clotet-Codina, M. Armand-Ugon, B. Grig-orov, D. Muriaux, B. Clotet, J. L. Darlix, and J. A. Este. 2005. Inhibition ofcoreceptor-independent cell-to-cell human immunodeficiency virus type 1transmission by a CD4-immunoglobulin G2 fusion protein. Antimicrob.Agents Chemother. 49:4296–4304.

4. Bosch, B., B. Grigorov, J. Senserrich, B. Clotet, J. L. Darlix, D. Muriaux, andJ. A. Este. 2008. A clathrin-dynamin-dependent endocytic pathway for the up-take of HIV-1 by direct T cell-T cell transmission. Antivir. Res. 80:185–193.

5. Burton, D. R., J. Pyati, R. Koduri, S. J. Sharp, G. B. Thornton, P. W. Parren,L. S. Sawyer, R. M. Hendry, N. Dunlop, P. L. Nara, et al. 1994. Efficientneutralization of primary isolates of HIV-1 by a recombinant human mono-clonal antibody. Science 266:1024–1027.

6. Chen, P., W. Hubner, M. A. Spinelli, and B. K. Chen. 2007. Predominantmode of HIV transfer between T cells is mediated by sustainedEnv-dependent neutralization-resistant virological synapses. J. Virol. 81:12582–12595.

7. Chenine, A., Q. J. Sattentau, and M. Moulard. 2000. Selective HIV-1-induced downmodulation of CD4 and coreceptors. Arch. Virol. 145:455–471.

8. Dimitrov, D. S., R. L. Willey, H. Sato, L. J. Chang, R. Blumenthal, and M. A.Martin. 1993. Quantitation of human immunodeficiency virus type 1 infec-tion kinetics. J. Virol. 67:2182–2190.

9. Favoreel, H. W., G. Van Minnebruggen, G. R. Van de Walle, J. Ficinska, andH. J. Nauwynck. 2006. Herpesvirus interference with virus-specific antibod-ies: bridging antibodies, internalizing antibodies, and hiding from antibodies.Vet. Microbiol. 113:257–263.

10. Ganesh, L., K. Leung, K. Lore, R. Levin, A. Panet, O. Schwartz, R. A. Koup,and G. J. Nabel. 2004. Infection of specific dendritic cells by CCR5-tropichuman immunodeficiency virus type 1 promotes cell-mediated transmissionof virus resistant to broadly neutralizing antibodies. J. Virol. 78:11980–11987.

11. Groot, F., S. Welsch, and Q. J. Sattentau. 2008. Efficient HIV-1 transmissionfrom macrophages to T cells across transient virological synapses. Blood111:4660–4663.

12. Gupta, P., R. Balachandran, M. Ho, A. Enrico, and C. Rinaldo. 1989. Cell-to-cell transmission of human immunodeficiency virus type 1 in the presenceof azidothymidine and neutralizing antibody. J. Virol. 63:2361–2365.

13. Haase, A. 1999. Population biology of HIV-1 infection: viral and CD4�

T-cell demographics and dynamics in lymphatic tissue. Annu. Rev. Immunol.17:625–656.

14. Healey, D., L. Dianda, J. P. Moore, J. S. McDougal, M. J. Moore, P. Estess,D. Buck, P. D. Kwong, P. C. L. Beverley, and Q. J. Sattentau. 1990. Novelanti-CD4 monoclonal antibodies separate human immunodeficiency virusinfection and fusion of CD4� cells from virus binding. J. Exp. Med. 172:1233–1242.

15. Hubner, W., G. P. McNerney, P. Chen, B. M. Dale, R. E. Gordon, F. Y.Chuang, X. D. Li, D. M. Asmuth, T. Huser, and B. K. Chen. 2009. Quanti-tative 3D video microscopy of HIV transfer across T-cell virological syn-apses. Science 323:1743–1747.

16. Johnson, D., and M. Huber. 2002. Directed egress of animal viruses pro-motes cell-to-cell spread. J. Virol. 76:1–8.

17. Jolly, C., K. Kashefi, M. Hollinshead, and Q. J. Sattentau. 2004. HIV-1 cellto cell transfer across an Env-induced, actin-dependent synapse. J. Exp.Med. 199:283–293.

18. Jolly, C., I. Mitar, and Q. J. Sattentau. 2007. Adhesion molecule interactionsfacilitate human immunodeficiency virus type-1-induced virological synapseformation between T cells. J. Virol. 81:13916–13921.

19. Jolly, C., I. Mitar, and Q. J. Sattentau. 2007. Requirement for an intactT-cell actin and tubulin cytoskeleton for efficient assembly and spread ofhuman immunodeficiency virus type 1. J. Virol. 81:5547–5560.

20. kremer, J., D. Mastronarde, and M. J. R. 1996. Computer visualization ofthree-dimensional image data using IMOD. J. Struct. Biol. 116:71–76.

21. Kuhmann, S. E., E. Platt, S. Kozak, and D. Kabat. 2000. Cooperation ofmultiple CCR5 coreceptors is required for infections by human immunode-ficiency virus type 1. J. Virol. 74:7005–7015.

22. Law, M., R. Hollinshead, and G. L. Smith. 2002. Antibody-sensitive andantibody-resistant cell-to-cell spread by vaccinia virus: role of the A33Rprotein in antibody-resistant spread. J. Gen. Virol. 83:209–222.

23. Levy, J. A. 2009. HIV pathogenesis: 25 years of progress and persistentchallenges. AIDS 23:147–160.

24. Majorovits, E., M. Nejmeddine, Y. Tanaka, G. P. Taylor, S. D. Fuller, andC. R. Bangham. 2008. Human T-lymphotropic virus-1 visualized at the viro-logical synapse by electron tomography. PLoS One 3:e2251.

25. Martin, N., and Q. Sattentau. 2009. Cell-to-cell HIV-1 spread and itsimplications for immune evasion. Curr. Opin. HIV AIDS 4:143–149.

26. Massanella, M., I. Puigdomenech, C. Cabrera, M. T. Fernandez-Figueras, A. Aucher, G. Gaibelet, D. Hudrisier, E. Garcia, M. Bofill, B.Clotet, and J. Blanco. 2009. Anti-gp41 antibodies fail to block earlyevents of virological synapses but inhibit HIV spread between T cells.AIDS 23:183–188.

27. Mastronarde, D. 2005. Automated electron microscope tomography us-ing robust prediction of specimen movements. J. Struct. Biol. 152:36–51.

28. Matthews, T., M. Salgo, M. Greenberg, J. Chung, R. DeMasi, and D. Bo-lognesi. 2004. Enfuvirtide: the first therapy to inhibit the entry of HIV-1 intohost CD4 lymphocytes. Nat. Rev. Drug Discov. 3:215–225.

29. Moore, J. P., S. G. Kitchen, P. Pugach, and J. A. Zack. 2004. The CCR5 andCXCR4 coreceptors–central to understanding the transmission and patho-genesis of human immunodeficiency virus type 1 infection. AIDS Res. Hum.Retrovir. 20:111–126.

30. Muster, T., F. Steindl, M. Purtscher, A. Trkola, A. Klima, G. Himmler, F.Ruker, and H. Katinger. 1993. A conserved neutralizing epitope on gp41 ofhuman immunodeficiency virus type 1. J. Virol. 67:6642–6647.

31. Pettersen, E., T. Goddard, C. Huang, G. Couch, D. Greenblatt, E. Meng, andT. Ferrin. 2004. UCSF Cimera: a visualization system for exploratory re-search and analysis. J. Comput. Biol. 25:1605–1612.

32. Phillips, D. 1994. The role of cell-to-cell transmission in HIV infection.AIDS 8:719–731.

33. Rudnicka, D., J. Feldmann, F. Porrot, S. Wietgrefe, S. Guadagnini, M. C.Prevost, J. Estaquier, A. T. Haase, N. Sol-Foulon, and O. Schwartz. 2009.Simultaneous cell-to-cell transmission of human immunodeficiency virus tomultiple targets through polysynapses. J. Virol. 83:6234–6246.

34. Ruggiero, E., R. Bona, C. Muratori, and M. Federico. 2008. Virologicalconsequences of early events following cell-cell contact between humanimmunodeficiency virus type 1-infected and uninfected CD4� cells. J. Virol.82:7773–7789.

35. Sabatos, C. A., J. Doh, S. Chakravarti, R. S. Friedman, P. G. Pandurangi,A. J. Tooley, and M. F. Krummel. 2008. A synaptic basis for paracrineinterleukin-2 signaling during homotypic T-cell interaction. Immunity. 29:238–248.

36. Sachdev, D. D., B. Zerhouni-Layachi, M. Ortigoza, A. T. Profy, M. Tuen,C. E. Hioe, and M. E. Klotman. 2009. The differential binding and activity ofPRO 2000 against diverse HIV-1 envelopes. J. Acquir. Immune Defic. Syndr.51:125–129.

37. Sanders, R. W., M. Venturi, L. Schiffner, R. Kalyanaraman, H. Katinger,K. O. Lloyd, P. D. Kwong, and J. P. Moore. 2002. The mannose-dependentepitope for neutralizing antibody 2G12 on human immunodeficiency virustype 1 glycoprotein gp120. J. Virol. 76:7293–7305.

38. Sato, H., J. Orenstein, D. Dimitrov, and M. Martin. 1992. Cell-cell spread ofHIV-1 occurs within minutes and may not involve the participation of virusparticles. Virology 186:712–724.

39. Sattentau, Q. 2008. Avoiding the void: cell-to-cell spread of human viruses.Nat. Rev. Microbiol. 6:815–826.

40. Scanlan, C. N., R. Pantophlet, M. R. Wormald, E. Ollmann Saphire, R.Stanfield, I. A. Wilson, H. Katinger, R. A. Dwek, P. M. Rudd, and D. R.Burton. 2002. The broadly neutralizing anti-human immunodeficiency virustype 1 antibody 2G12 recognizes a cluster of �132 mannose residues on theouter face of gp120. J. Virol. 76:7306–7321.

41. Sims, T. N., and M. L. Dustin. 2002. The immunological synapse: integrinstake the stage. Immunol. Rev. 186:100–117.

42. Sourisseau, M., N. Sol-Foulon, F. Porrot, F. Blanchet, and O. Schwartz.2007. Inefficient human immunodeficiency virus replication in mobile lym-phocytes. J. Virol. 81:1000–1012.

43. Sowinski, S., C. Jolly, O. Berninghausen, M. A. Purbhoo, A. Chauveau, K.Kohler, S. Oddos, P. Eissmann, F. M. Brodsky, C. Hopkins, B. Onfelt, Q.Sattentau, and D. M. Davis. 2008. Membrane nanotubes physically connectT cells over long distances presenting a novel route for HIV-1 transmission.Nat. Cell Biol. 10:211–219.

44. Timpe, J. M., Z. Stamataki, A. Jennings, K. Hu, M. J. Farquhar, H. J.Harris, A. Schwarz, I. Desombere, G. L. Roels, P. Balfe, and J. A. McKeat-ing. 2008. Hepatitis C virus cell-cell transmission in hepatoma cells in thepresence of neutralizing antibodies. Hepatology 47:17–24.

45. Trepel, F. 1974. Number and distribution of lymphocytes in man: a criticalanalysis. Klin. Wochenschr. 52:511–515.

46. van Montfort, T., A. A. Nabatov, T. B. Geijtenbeek, G. Pollakis, and W. A.Paxton. 2007. Efficient capture of antibody neutralized HIV-1 by cells ex-pressing DC-SIGN and transfer to CD4� T lymphocytes. J. Immunol. 178:3177–3185.



http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org/

virological synapse-mediated spread of human immunodeficiency

Documents